M0.3 Benchmark: OPFS sync-access-handle throughput at scale

[melvincarvalho]

Part of #1 (M0).

Goal

Measure read/write throughput via createSyncAccessHandle() in a Web Worker, in a block-file layout, scaling toward tens of GB.

Benchmark / acceptance

• Sequential write and read MB/s; random-read latency for UTXO-style access.
• Behaviour and stability approaching tens of GB and many files.
• Persistence across reloads with navigator.storage.persist(); note any quota or eviction.
• A go/no-go read on whether OPFS can plausibly back a ~650 GB store (Tier 2).

Notes

OPFS needs no special HTTP headers, so it works on gh-pages. This is the least battle-tested pillar at scale; surface the sharp edges early.